How to Reduce Noise in Low-Light Photos: A Complete Denoising Guide for 2026
High ISO isn't the compromise it used to be
For most of digital photography's history, "shoot at a high ISO" came with an asterisk. You'd get the exposure you needed in dim light, but you'd pay for it with grainy, mottled shadows and color noise that turned smooth skin and skies into a mess of speckled pixels. Photographers built entire workflows around avoiding high ISO: wider apertures, slower shutter speeds propped up on tripods, flash when nothing else worked.
AI-powered denoise tools have quietly rewritten that trade-off. Lightroom's Denoise feature, Photoshop's neural filter based noise reduction, and dedicated tools like Topaz and DxO PureRAW can now recover detail from files that would have been unusable five years ago. ISO 12,800 files that used to look like sandpaper can come out clean enough for a full-page print.
This guide walks through what's actually happening when these tools remove noise, how to use the built-in options in Lightroom and Photoshop, when a dedicated third-party tool is worth the money, and how to avoid the over-processed, plastic look that bad denoising produces.
As an Amazon Associate, ExifGrabber earns from qualifying purchases.
Why photos get noisy in the first place
Noise comes from two main sources. The first is signal-to-noise ratio: in low light, your sensor is capturing relatively few photons, so the electrical noise inherent in the sensor and processing chain makes up a bigger proportion of the final signal. This is why noise gets dramatically worse as ISO climbs, you're amplifying a weak signal, and amplifying the noise along with it.
The second source is heat and sensor design. Smaller sensors (like those in phones and compact cameras) pack photosites closer together, which generally means more noise at a given ISO than a larger full-frame or APS-C sensor. Long exposures add their own thermal noise as the sensor heats up during extended capture, which is why star trail and astrophotography images often need dedicated noise reduction techniques like dark frame subtraction on top of standard denoising.
There are two visible types of noise you'll be fighting: luminance noise, which looks like fine grain or texture in flat areas, and chroma (color) noise, which shows up as random red, green, or blue speckles, especially in shadows. Chroma noise is generally considered more objectionable and is usually removed more aggressively, while a small amount of luminance noise can even be left in intentionally to preserve a natural, filmic texture.
Lightroom's AI Denoise
Adobe's AI Denoise feature, built into Lightroom Classic and Camera Raw, is trained on a large dataset of noisy and clean image pairs rather than relying on the traditional luminance and color noise sliders alone. The difference from the old approach is significant: traditional noise reduction essentially blurs noise away, softening fine detail in the process. AI Denoise tries to recognize the difference between noise and real detail, so it can remove one while mostly preserving the other.
How to use it:
- Open your RAW file in the Develop module (AI Denoise currently only works on RAW files, not JPEGs)
- Go to the Detail panel and click the Denoise button
- A preview window opens showing a before and after crop. Zoom into a detailed, noisy area of the frame
- Adjust the Amount slider, higher values apply more aggressive noise removal
- Click Enhance to process. This creates a new DNG file alongside your original, since the process is computationally heavy and not applied in real time like other sliders
The output DNG becomes your new working file, so all subsequent edits happen on the denoised version. This is worth knowing because it means Lightroom AI Denoise is a step you generally want to run early in your editing process, before doing detailed local adjustments.
Photoshop's Camera Raw denoise
Photoshop shares the same underlying Camera Raw engine as Lightroom, so if you're editing a RAW file through the Camera Raw filter or the Camera Raw import dialog, you have access to the same neural network based denoising. Open your file, go to Filter, then Camera Raw Filter, and you'll find the Denoise controls in the same place as in Lightroom's Detail panel.
For JPEGs and TIFFs that have already been flattened, Photoshop's traditional Reduce Noise filter (under Filter, then Noise) is still available, though it uses the older, less selective approach and tends to soften detail more aggressively than the AI-based option. If you're working from a RAW file, always denoise before converting to a flattened format, since you lose the option for AI-based denoising once the file becomes a JPEG.
When to reach for dedicated denoise software
Adobe's built-in tools handle the vast majority of noise reduction needs well, but there are situations where a dedicated third-party tool pulls ahead, usually in extreme high ISO files, astrophotography, or when you need noise reduction as a lightweight preprocessing step before your normal editing workflow rather than baked into Lightroom's DNG conversion.
Topaz Photo AI / DeNoise AI remains one of the most capable standalone denoise engines on the market, particularly for extreme low light and wildlife shooters working with heavily cropped telephoto files. Topaz has moved to a subscription model for its full Photo AI suite at $399 a year, though the standalone DeNoise AI tool is available separately for $79.99 a year for photographers who only need noise reduction without the sharpening and upscaling features bundled into the full suite.
DxO PureRAW takes a different approach. Rather than editing inside your normal workflow, it's a preprocessing tool: you run your RAW files through it first, it applies DxO's DeepPRIME denoising algorithm along with lens corrections, and outputs clean DNG files that you then import into Lightroom or Capture One as normal. The current version starts at $139.99 for new licenses, with a $89.99 upgrade path from earlier versions. It's popular with photographers who want denoising handled once, automatically, in a batch, rather than a manual step per image.
ON1 NoNoise AI is the budget-friendly option in this category, priced at $49.99 with a perpetual license rather than a subscription. It doesn't quite match Topaz or DxO on the most extreme high ISO files, but for typical low light shooting it produces very solid results and the one-time cost makes it appealing for photographers who don't want another recurring subscription.
A practical denoise workflow
Here's a sequence that works well for most low light and high ISO photos:
- Shoot RAW. Denoising a JPEG loses far more detail than denoising a RAW file, since the compression has already discarded information the denoise algorithm could otherwise use.
- Expose to the right whenever possible. A slightly brighter exposure at capture time, pulled back down in editing, generally produces less visible noise than an underexposed shot pushed up later. Check your histogram, not just your LCD preview, to confirm you're not underexposing out of habit.
- Apply AI Denoise first, before other edits. Whether you're using Lightroom's built-in tool or a dedicated app like Topaz or DxO PureRAW, run denoising near the start of your workflow so the rest of your edits (sharpening, local adjustments, color grading) are working with clean data.
- Zoom to 100% and check real detail, not just noise. The biggest mistake in denoising is judging the result at a small preview size. Noise that looks perfectly handled zoomed out can reveal waxy, over-smoothed skin or lost texture in hair and foliage once you zoom in.
- Don't chase zero noise. A very small amount of fine luminance noise, retained deliberately, often looks more natural than a completely noise-free image, which can start to look artificial or overly digital, especially in skin tones.
Avoiding the "AI plastic" look
The most common complaint about aggressive AI denoising is that it makes photos look artificial, an effect often described as plastic or waxy skin, especially in portraits and wildlife shots with fur or feather detail. This happens when the denoise amount is pushed too high, smoothing away texture the algorithm couldn't distinguish from noise.
The fix is almost always to back off the Amount slider and accept a small amount of remaining noise rather than chasing a completely clean file. Most tools, including Lightroom's AI Denoise and Topaz, let you preview at 100% before committing, so use that preview aggressively. If a client or print use case genuinely requires maximum noise removal, consider applying it selectively with a mask rather than globally, so detailed areas like eyes, hair, and fabric texture stay sharper than large, flat areas like skies and out-of-focus backgrounds.
Checking what ISO you actually shot at
If you're building a personal library of "how noisy is too noisy" for your specific camera, it helps to know exactly what settings produced a given result. Our free EXIF viewer reads the ISO, shutter speed, and aperture embedded in a photo's metadata directly in your browser, with nothing uploaded to a server. It's a fast way to confirm whether a noisy shot was actually pushed to ISO 12,800, or whether the noise is coming from underexposure at a more moderate ISO that got brightened significantly in editing, since the fix for each of those problems is different.
This is also useful when comparing gear. If you're deciding whether a camera's high ISO performance is good enough for your low light work, pulling the EXIF data from sample images online (many review sites embed full-resolution samples) lets you confirm exactly what settings and camera produced a given noise level, rather than trusting a headline claim.
Denoise and astrophotography
Astrophotography pushes noise reduction further than almost any other genre, since long exposures at high ISO in the dark are, by definition, the worst-case scenario for sensor noise. Most astro workflows use dedicated techniques on top of standard denoise tools, including dark frame subtraction (capturing exposures with the lens capped to record only the sensor's thermal noise pattern, then subtracting it) and stacking multiple exposures in software to average out random noise while reinforcing consistent detail like stars and nebulae. If you're getting into night sky photography, our guide to how to photograph star trails covers exposure settings that minimize noise before you even reach for a denoise tool.
Comparing your options at a glance
| Tool | Price | Best for | Workflow style |
|---|---|---|---|
| Lightroom AI Denoise | Included with Creative Cloud photography plans | Everyday high ISO files, RAW only | Built into Develop module |
| Photoshop Camera Raw Denoise | Included with Creative Cloud | RAW files opened via Camera Raw Filter | Filter-based, same engine as Lightroom |
| Topaz DeNoise AI | $79.99/year (standalone) | Extreme low light, wildlife, heavy crops | Standalone app or plugin |
| DxO PureRAW | $139.99 new, $89.99 upgrade | Batch preprocessing before editing | Standalone, outputs clean DNGs |
| ON1 NoNoise AI | $49.99 perpetual license | Budget-conscious photographers, typical low light | Standalone or plugin, one-time cost |
Frequently asked questions
Does denoising work on JPEGs? Yes, but with meaningfully worse results than RAW files. JPEG compression has already thrown away image data that AI denoise algorithms use to distinguish real detail from noise. If you have the option, always denoise from the original RAW file and export a JPEG afterward, rather than denoising a JPEG that's already been compressed.
Will denoising slow down my editing workflow? AI-based denoising is computationally heavy compared to a simple slider adjustment, so yes, expect processing to take anywhere from a few seconds to over a minute per image depending on your computer's GPU and the resolution of the file. Batch processing a large shoot through DxO PureRAW or Lightroom's Denoise before you start detailed editing is usually more efficient than denoising individual images as you go.
Can I use more than one denoise tool on the same image? It's possible, but generally not worth it. Running Lightroom's AI Denoise and then also pushing an already-denoised file through Topaz tends to produce diminishing returns and can start introducing the plastic, over-smoothed look faster than either tool used on its own. Pick one primary denoise step in your workflow.
Is a full-frame camera always less noisy than a crop sensor? Generally yes, at the same ISO, a larger sensor with larger individual photosites tends to gather more light per pixel and produce a cleaner signal, but sensor generation and processing matter too. A newer APS-C or Micro Four Thirds sensor can outperform an older full-frame sensor at high ISO. If you're comparing gear, check sample images and their embedded ISO data directly rather than assuming sensor size alone tells the whole story.
The bottom line
Modern AI denoising has meaningfully changed what's usable in low light. A file shot at ISO 6400 or higher, once reserved for small web use at best, can now often hold up to a reasonably large print after a pass through Lightroom's Denoise, Topaz, or DxO PureRAW. The built-in Lightroom and Photoshop tools cover most everyday needs at no extra cost, and they're the right starting point for almost everyone. Dedicated software earns its keep for wildlife, astrophotography, and other genres where you're routinely working at the edges of what your sensor can handle. Whichever tool you use, the same rule applies: shoot RAW, expose well, and don't chase a perfectly clean image at the expense of real detail.